Abstract
In this paper I report the discovery of neurons which showed a neural correlate with ongoing fluctuations of Bitcoin and Ethereum prices at the time of the recording. I used the publicly available dataset of Neuropixel recordings by the Allen Institute to correlate the firing rate of single neurons with cryptocurrency price. Out of ∼40.000 recorded single neurons, ∼70% showed a significant correlation with Bitcoin or Ethereum prices. Even when using the conservative Bonferroni correction for multiple comparisons, ∼35% of neurons showed a significant correlation, which is well above the expected false positive rate of 5%. These results were due to ‘nonsense correlations’: when correlating two signals which both evolve slowly over time, the chances of finding a significant correlation between the two are much higher than when comparing signals which lack this property.
When this paper first popped into my Twitter feed I only got to see the abstract above and thought WTF? But now that I have had time to read through the entire thing and seen the complete title –Neurons in the mouse brain correlate with cryptocurrency price: a cautionary tale – it makes sense as shown in the conclusion below.
When analysing signals that slowly evolve over time,like neural activity, one should take the utmost care to avoid the pitfall of ‘nonsense correlations’ [1]. Although this issue is widely discussed in other scientific fields, its importance is only recently gaining traction in systems neuroscience. This paper serves as a cautionary tale that the potential confound of nonsense correlations is to be taken seriously. When not properly controlled for, it can lead to the misleading conclusion that 70% of neurons in the mouse brain encode cryptocurrency prices.
As might be expected mice neurons in no way encode cryptocurrency prices as the author somewhat tongue in cheek states since mice lack the ability to read and interpret complex financial data then any correlation must have simply evolved as a statistical quirk. Whilst this may seem like a tale for data scientists to take care in the conclusions they derive from large data sets it is also a warning shot for traders. I will state categorically that most trading indicators be they technical fundamental or whatever add little if anything to the trading process beyong being a form of hand-holding and the further you get away from price as your prime tool the worse this problem becomes. Yet traders swear by all sorts of strange nonscientific bollocks such as astrology and Gann despite there being nothing to support them beyond religious fervour and the occasional incident of correlation. As I am constantly telling people – if you draw enough lines on a chart enough times then sooner or later one of them will be right. Unfortunately, most traders will interpret this as a moment of genius on their behalf and believe incorrectly that their particular tool or interpretative measure can predict the future.
Someone has to win the lottery each week – just because you managed to pick the right numbers doe not mean that you can predict the future or that there is anything special about your numbers despite the overwhelming tendencies to believe there is.
I am sad to read all this. Neurocorrelates of this cross-mammalian nature, to my knowledge, have not yet been subjected to rigorous peer reviewed studies. It is much easier to assert and support propositions of positive rather than negative correlations based on propositions of intuitive probability. I am seriously concerned that there is a little bit of conspiracy theory and flat earth philosophy embedded in the article. I hope none of your readers take this too seriously until the full weight of scientific investigation has been brought to bear. Don’t hold your breath. After all string theory and chaos theory seemed absurd initially. And then there is the ginger headed fool. I mean, how improbable do things need to be when they become truly accepted by those of the common ilk, like myself.
Unfortunately many will not get the author’s ironic take on correlations and the need to avoid junk correlations. I have already seen it quoted on my social media feeds where clearly someone has only either read the title or the first sentence of the abstract without reading the actual conclusion.